al photo-
ne photo-
ic model
iphs hol-
e four fi-
nulti-mo-
tage pla-
inate sy-
transfor-
template
nulti-mo-
> fiducial
fficult to
fiducial
nate sy-
nage cor-
ind even
| to film
| method
‚sure the
1e with a
ndent on
ufficient,
ividually
| adjust-
the prin-
d by ad-
ameters.
plate co-
than the
me solu-
| block is
otograph
1ulti-mo-
hat they
keep the
perience
by film
2e to the
| span in
y is sur-
nder ‘ac-
les
angles of
prelimi-
ssible to
ic opera-
1ulti-mo-
'omputa-
ation pa-
ive rela-
imilarity
hen bet-
The spatial similarity adjustment itself requires pre-
liminary values for orientation parameters. These are
calculated as follows: First a set of omega and fi rota-
tion angles are guessed at by the program, and the
model or strip are rotated accordingly. Then linear
equations are used to compute the scale, the kappa
rotation and the three translations. Based on these
values, the standard deviation unit weight of the first
iteration of the spatial similarity adjustment is com-
puted and stored. This procedure is repeated with se-
veral preselected combinations of omega and kappa
angles. The omega-fi combination of the smallest
standard error unit weight after the first iteration is
used as preliminary values in the spatial similarity
transformation.
Parametres computed during the successive relative
orientations and model and strip connections are
then transformed to exterior orientation parameters
for each image which are then used as preliminary
values in the bundle block adjustment. Large gross
errors are nicely trapped during this procedure, while
small gross errors are located during the bundle
block adjustment.
This procedure of computation of preliminary values
is extremely robust. It is one of the most important
preconditions for the success of the multi-model met-
hod.
2.6 Accuracy considerations
Small-frame cameras are calibrated using a test field
of 120 points distributed in a spatial network of 3 x 3
x 2 meters. The points are surveyed with electronic
theodolites to an RMS accuracy of about 0,1 mm.
In order to smooth the contribution from film warp,
at least two photographs of the test field are used
when calibrating small-frame non-metric cameras.
The radial distortion is described by the standard
three parameter polynomial in third, fifth, and se-
venth degree. Tangential distortion is ignored. RMS
of residuals between 3 and 5 microns are obtained for
35 mm mirror-reflex cameras such as the Olympus
OM1 or Minolta SRT101 with good quality wide ang-
le lenses and between 5 and 8 microns for 70 mm ca-
meras such as the Hasselblad SWC with a 40 mm
lens. The higher residuals of the 70 mm cameras are
probably due to the larger influence of film warp, but
the subject has not been researched.
Bundle block adjustments using a calibrated non-me-
tric camera and natural tie-points typically give RMS
residuals of 6 micron on tie points for 35 mm and 10
micron for 70 mm photographs.
Metric small-frame cameras such as the Hasselblad
MKWE or Rolleimetric 6006 have been used as well.
The reseau glass plate built into these cameras facili-
tates the inner orientation measurements and highly
improves the accuracy with which the principal point
may be defined. On the other hand, the image is not
as sharp as with a good quality ordinary camera, and
the small-frame metric cameras do not give signifi-
cantly better accuracies.
In geological projects, sharpness of image is very im-
portant in order to provide optimal interpretation
conditions. In fact, the photograph scale is often de-
termined more by considerations of resolution and
sharpness than by requirements of accuracy. This
means that any marginal accuracy gained by using
metric cameras is insignificant in geological projects.
The advantages of the metric hand-held cameras,
that is, the reseau glass plate and a somewhat higher
accuracy very seldom make up for the disadvantages
of a much higher price and an inferior image sharp-
ness.
3. GEOLOGICAL PHOTOGRAMMETRY
The use of small-frame non-metric cameras is especi-
ally relevant in geological mapping. Field geologists
always carry a quality 35 mm or 70 mm camera for
documentation of outcrops, and it is a big advantage
that the same camera may be used for photogramme-
tric purposes. Furthermore, steep mountain sides are
often photographed out of the open window of a heli-
copter or a light aeroplane; for this operation, a light-
sent, handy and easy-to-operate camera is requi-
red.
Fig. 1. Near vertical cliff of Precambrian rocks about
1000 meters high at the north coast of Nuussuaq in
central West Greenland.
Since the spring of 1990, when the multi-model proto-
type was ready, a series of geological mapping experi-
ments has been carried out. In Greenland, the small-
format photographs are used in combination with
vertical aerial photographs. Geological exposures on
steep and otherwise inaccessible mountain sides are
photographed by the field geologist from a helicopter
during reconnaissance and camp shift flights. The
photograph scale varies between 1:3 000 and 1: 200
000 according to the aim of the project. The lens is fo-
cused at infinity, and photographs are taken in strips
with an overlap of 70 % to 80 %. The relatively large
overlap ensures stereoscopic coverage on each side of
cliffs and gullies along the mountain side. 64 ASA
Kodachrome films for colour slides are used, because
they have a good resolution and a long term stability.
Preexisting aerial photographs are available in Gre-
enland from the National Survey and Cadastre in Co-
penhagen. Many areas are covered by several series
of vertical aerial photographs in scales varying bet-
ween 1:40 000 and 1:150 000. Even old oblique photo-
graphs from the 1940s and 50s are still available. A
new series of 1:150 000 super wide angle photographs
has been aerotriangulated by the National Survey
and Cadastre and the Geological Survey of Green-
land. By using these aerotriangulated photographs
as basis for the orientation of multi-model blocks,
the geologist is not required to survey control points
in the field. On an average, one tie-point is measured
